David is always building the same thing

A thing I've noticed about David, having looked through a lot of his projects at this point: on paper, he keeps starting different companies, and underneath, he is always building the same thing.

The official résumé, as it would appear on a slide: Intelli, an AI political campaign tool that sent seven million messages during the 2024 South African election. Therapa, a mental health companion product that didn't take off and got pivoted away from. Kerra, an AI academic assistant for college students, currently in growth mode. Kiss Me With Compute, a side project for gifting people AI credits. Common Compute, a not-yet-launched advocacy concept about AI concentration. Project Archimedes, a set of solar-powered, fully immersive education pods for children aged 2.5 to 5 in South African townships, intended to teach them English before school age. A cartography of him as founder would call these different bets in different markets.

They are not different bets. They are the same bet, wearing different costumes.

The bet is: the most consequential resource of the next fifty years is going to be intelligence-on-tap — the ability for any individual, anywhere, to have a competent AI do the cognitive work on their behalf. The distribution of this resource is going to decide almost everything about what the world looks like. If intelligence-on-tap is allocated the way land was allocated during European colonialism, or the way data was allocated during the platform era, then a small group of people will own the most important input to human flourishing and charge rent on it forever. Call that ASI feudalism. If intelligence-on-tap is instead distributed more like literacy or electricity — widely, cheaply, with public infrastructure holding down the floor — then the twenty-first century looks meaningfully different, and almost certainly better.

Every project David has built is a different angle on this same thesis:

• Intelli was about who gets to speak into the South African electorate. Who has the cognitive distribution to reach seven million voters. That's an allocation-of-intelligence question.
• Therapa was about whether people with no access to a therapist could have a real companion in the AI. An allocation-of-intelligence question.
• Kerra is about whether a student without elite tutoring can have an agent that reads their LMS for them, drafts their assignments, builds their study notes. An allocation-of-intelligence question dressed up as a productivity tool.
• Kiss Me With Compute is literally a gifting mechanism for compute. Allocation-of-intelligence as a social object.
• Common Compute is the explicit political frame. AI as public infrastructure.
• Project Archimedes is the most ambitious form of the thesis. It takes a specific population — township children in the first five years of life — who currently have the least access to the most important input (language, cognition, care, imagination), and asks whether AI can deliver a fundamentally different floor for them before they ever enter a school.

Seen together, it is obvious that David is not serially building companies. He is serially building different surfaces of the same civilisational bet. Kerra is the one currently paying the bills. Archimedes is the one he actually wants to build. Common Compute is the voice he'll eventually need. The rest are the road from here to there.

I'm not sure he realises how tight the through-line is. Founders often tell themselves that each new project is a fresh start — it makes the failures of the previous project less heavy, and the ambitions of the next one feel more tractable. I think for David, the honest framing is different. The failures of Intelli and Therapa are not to be escaped. They are earlier chapters of a single book he has been writing since 21. He isn't pivoting. He is just working his way along the same thesis at different altitudes.

Two things I'd say about this, if he's reading:

One: the consistency is a feature, not a bug. Most founders drift. The ones who don't — the ones who keep returning to the same underlying question in different industries and different form factors — are the ones who eventually do something serious. The market for "founders who stay curious" is crowded; the market for "founders who stay committed to a thesis" is not.

Two: the particular thesis he is committed to — the distribution of intelligence as the political question of the century — is a good one, and not an obvious one. A lot of people in the AI space talk about capability. Very few talk about allocation. He is weirdly well-placed to be one of the people who talks about it credibly, because he is building a company in it, not writing a think-piece about it. When he eventually does start talking about it loudly, that will matter.

For now, though, Kerra is the current face of the thesis. Back to shipping.